OBJECTIVE To examine trends over time in the rates of admission to hospital for accidents of preschool children and to study patterns of repeated admissions for accidents in these children. METHODS Analysis of linked, routine abstracts of hospital inpatient records for accidents. METHODS Six districts in the Oxford Regional Health Authority covered by the Oxford record linkage study. METHODS Records for 19,427 children aged 5 years and under at the time of first recorded admission to hospital. METHODS Number of admissions to hospital. RESULTS Records were analysed in three groups: person based annual admission rates were calculated for each calendar year; each child's first recorded admission in 1976-85 was identified, and the child's record was followed up by linkage for one year from that admission; each child's first recorded admission in 1976-81 was identified and followed up for five years. Overall, 19,427 children from an average annual resident population of 163,000 children in 1976-86 had 20,657 admissions for accidents before they were 6 years of age. Of these admissions 13,983 were for injuries, 5717 for poisonings, and 957 for burns. Admission rates declined after 1976 for poisoning, but no substantial changes over time were found in admission rates for injuries or burns. A total of 17,724 children were followed up for one year and 10,889 for five years; 470 (2.6%) of the children who were followed up for one year and 926 (8.5%) of those followed up for five years had at least one further admission for an accident. Of those followed up for one year the 4 and 5 year old children were least likely and those under 1 and 1 year old were most likely to have a further admission for an accident. The number of children who had more than one accident was greater than would be expected if accidents were random occurrences. Those who had a poisoning at first admission were more likely to have another poisoning than an injury or burn; and those who had a burn at first admission were more likely to have another burn. CONCLUSIONS Hospital admissions for accidents in children are common: on average 1 child in 88 in this population was admitted each year. Multiple admissions are uncommon but none the less occur more often than would be expected by chance.